First attempt at fixing a crash that occurs when we get into WebFrameLoaderCl...

First attempt at fixing a crash that occurs when we get into WebFrameLoaderClient::dispatchDidChangeLocationWithinPage()
with an empty redirect chain.

BUG=22205
TEST=none



Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=30496

[darin (slow to review), 2009-10-23 20:38:14]

http://codereview.chromium.org/326015/diff/1/2
File webkit/glue/webframeloaderclient_impl.cc (right):

http://codereview.chromium.org/326015/diff/1/2#newcode498
Line 498: KURL chain_end = KURL();
nit: just do "KURL chain_end;" since that is equivalent.

[darin (slow to review), 2009-10-23 20:38:29]

also, this needs a test of some sort.  a layout test would be fine.

[Nate Chapin, 2009-10-23 20:49:52]

I'll see what I can do, though thus far I've had trouble finding a suitable
reduction.

On Fri, Oct 23, 2009 at 1:38 PM, <darin@chromium.org> wrote:

> also, this needs a test of some sort.  a layout test would be fine.
>
>
> http://codereview.chromium.org/326015
>

[darin (slow to review), 2009-10-23 20:52:37]

okay.... so this is a patch based on crash analysis data?
-darin

On Fri, Oct 23, 2009 at 1:49 PM, Nate Chapin <japhet@chromium.org> wrote:

> I'll see what I can do, though thus far I've had trouble finding a suitable
> reduction.
>
>
> On Fri, Oct 23, 2009 at 1:38 PM, <darin@chromium.org> wrote:
>
>> also, this needs a test of some sort.  a layout test would be fine.
>>
>>
>> http://codereview.chromium.org/326015
>>
>
>

[Nate Chapin, 2009-10-29 17:14:24]

On 2009/10/23 20:52:37, darin wrote:
> okay.... so this is a patch based on crash analysis data?
> -darin
> 
> On Fri, Oct 23, 2009 at 1:49 PM, Nate Chapin <mailto:japhet@chromium.org>
wrote:
> 
> > I'll see what I can do, though thus far I've had trouble finding a suitable
> > reduction.
> >
> >
> > On Fri, Oct 23, 2009 at 1:38 PM, <mailto:darin@chromium.org> wrote:
> >
> >> also, this needs a test of some sort.  a layout test would be fine.
> >>
> >>
> >> http://codereview.chromium.org/326015
> >>
> >
> >
> 

Fixed the nit.

Crash dump can be found on the bug (http://crbug.com/22205).  I haven't found a
suitable reduction yet.

I don't fully understand how chain_end gets used in the rest of the fucntion. 
Specifically, in the initialization of was_client_redirect, is there any reason
I should a condition like "!chain_end.isEmpty()" to the and clause?

Is there anything else I should be doing to try to help us determine how we get
into the state where this crash is possible?

[darin (slow to review), 2009-10-29 17:50:02]

LGTM